Our Services in Detail

Name: Konsultacje i wdrożenia AI dla firm
Price: 149 PLN

We build four categories of AI systems. Each has a defined scope, a defined deliverable, and a defined price.

The categories are not arbitrary. They reflect four levels of complexity, and each level requires a different amount of planning, building, and testing. Choosing the right tier means matching the solution to the problem — not overshooting with a system when an assistant would do, and not underpowering with an assistant when a system is needed.

Tier 1 — AI Guide (149 PLN)

What it is: A structured consultation for founders and operators who want to understand whether AI applies to their business before spending money on a build.

This is not a sales call. It is an honest assessment of your situation by someone who builds AI systems for a living and has no incentive to oversell you.

How this differs from free Discovery: Discovery sessions (free) are for clients who already know they want to explore a Tier 2+ build. The AI Guide is for people who are not sure whether AI applies to their business at all and want a standalone answer with no commitment to a larger project.

What Happens During the Session

The AI Guide follows a clear three-step format:

Step 1: We review your current processes together (45 minutes). You walk us through your daily and weekly operations. Not the org chart — the actual work. What tasks happen repeatedly, who does them, how long they take, and what tools are involved. We ask specific questions. We push for numbers.

Step 2: We identify 3-5 specific areas where AI applies (45 minutes). Based on what you have told us, we highlight the processes that are good candidates for AI — and equally important, the ones that are not. We explain the reasoning for each. You will understand not just "what" but "why."

Step 3: You receive a written assessment within 48 hours. Not a generic report. A specific, actionable document tailored to your business.

What the Written Assessment Looks Like

The assessment contains specific action items with concrete estimates. Here is what actual entries look like:

›
"Email classification and routing: estimated 4 hours/week saved. Your team currently reads every incoming email, categorizes it manually, and forwards it. An AI assistant can handle this with 90%+ accuracy for standard categories. Recommended: Tier 2 assistant with Gmail integration. Estimated cost: 2,000-3,000 PLN."
›
"Report generation: your current process is too unstructured for automation. Reports are assembled differently each time depending on who is doing it, with no consistent template or data source. Recommended: standardize the report template and data sources first. Revisit AI automation in 3 months after the process is consistent."
›
"Client onboarding: strong candidate for Tier 3 automation. The process is clearly defined (intake form → document collection → CRM entry → welcome email → task assignment) and happens 15-20 times per month. Estimated cost: 5,000-7,000 PLN. Estimated time saved: 8-10 hours/month."

Each item gets a clear verdict: worth automating now, worth automating later (with prerequisites), or not worth automating.

Who It is For

Business owners and operators who have heard about AI, are skeptical of the hype, and want a grounded second opinion on whether it applies to their work. People who would rather spend 149 PLN on an honest assessment than 10,000 PLN on a project that might not be the right fit.

Who It is NOT For

People looking for a generic introduction to AI. We do not do "what is ChatGPT and how does it work." If you want that, there are free resources everywhere. The AI Guide is about your business, your processes, your specific situation.

Also not for people who have already decided to build and just want a quote. If you know what you need, skip to the consultation and go directly to Discovery for a Tier 2+ project.

What People Usually Discover

Most AI Guide clients discover two things. First, that AI applies to fewer processes than they expected — but the ones it does apply to are high-impact. Second, that some of their biggest time sinks require process standardization before AI can help. The written assessment gives them a concrete plan for both: what to automate now and what to fix first.

Note on Pricing

The 149 PLN you pay for the AI Guide is deducted from any Tier 2 or higher project started within 60 days. If the Guide leads to a project, the consultation was effectively free.

Tier 2 — AI Assistant (from 1,500 PLN)

What it is: A custom AI assistant integrated into one specific workflow. It knows your business, your documents, your tone, and your rules. It handles one defined task, and it handles it well.

Three Use Case Scenarios in Detail

1. Client Proposal Drafting

Your current process: A client fills out an inquiry form or sends an email. Someone on your team reads it, opens a Word template, manually writes a proposal matching your style and pricing, attaches relevant documents, and sends it. Takes 30-60 minutes per proposal. You do this 10-15 times per week.

With an AI assistant: The client submits an intake form. The AI reads the form data, pulls relevant pricing from your rate sheet, generates a proposal draft matching your template and tone, and presents it to a team member for review. The team member spends 5 minutes checking it, makes minor edits, and sends it. Time per proposal: 10 minutes instead of 45.

Over a month, that is roughly 30-40 hours saved. The assistant pays for itself in the first week.

2. Internal Knowledge Base

Your current process: Employees have questions about policies, procedures, product details, or client history. They ask a colleague, search through shared drives, or email the one person who remembers. The person who knows the answer gets interrupted 15 times a day. New employees take months to become self-sufficient.

With an AI assistant: Employees ask questions in a chat interface. The AI retrieves answers from your documents, policies, and procedures. It answers in your format, references the source document, and flags when it is not confident. The knowledgeable employee stops being the bottleneck. New employees get answers in minutes instead of hours.

The knowledge base is not a search engine. It understands context. "What is our return policy for enterprise clients who signed before 2024?" gets a specific answer, not a list of documents to read.

3. Email Classification and Routing

Your current process: Someone reads every incoming email to the company inbox. They decide: is this a sales inquiry, a support request, an urgent complaint, spam, or something else? They forward it to the right person, sometimes with a brief summary. Takes 3-5 hours per week for a company receiving 100-200 emails per day.

With an AI assistant: Incoming emails are automatically classified into your defined categories. The AI identifies urgency, routes to the right person, and drafts a brief summary. It can also draft a response for the team member to review before sending. For clearly routine emails (meeting confirmations, auto-replies, newsletters), it handles them automatically.

Human review stays in the loop for anything ambiguous or high-stakes. The AI handles the 80% that is routine, your team handles the 20% that needs judgment.

What the Deliverable Looks Like

When the project is done, you receive:

›Working assistant accessible via web interface or API (depending on your needs)
›System prompt and knowledge base — all configuration files, fully documented
›Integration with one external tool (email, CRM, document storage, calendar — your choice)
›Handoff documentation — how it works, how to maintain it, how to update it
›30-day bug support — we fix anything that does not work as specified. This covers bugs (system does not do what the spec says) not feature requests (system does not do something that was not in the spec)

Everything is delivered as a GitHub repository. You own the code. Full change history, all configuration files, complete documentation.

What Tier 2 Does Not Include

To be explicit: a Tier 2 assistant handles one workflow with one integration. It does not handle multiple unrelated tasks, connect to multiple external tools, or run without any human oversight. If you need those things, you are looking at Tier 3 or Tier 4.

Also not included: ongoing prompt tuning, new feature development, or adapting the assistant to new workflows after delivery. The 30-day support covers bugs — the system not doing what the spec says. Changes to what the system does are new work, quoted separately.

Timeline

2-4 weeks from signed specification. The simpler the integration, the shorter the timeline. Most Tier 2 projects land at 3 weeks.

Tier 3 — Process Automation (from 4,000 PLN)

What it is: An automated pipeline that removes a specific recurring manual process from your operations. Unlike Tier 2 (which handles one task with human involvement), Tier 3 automates an entire process end-to-end, from trigger to delivery.

How a Pipeline Works

Every Tier 3 project follows this general pattern:

Trigger → Intake → AI Processing → Validation → Output → Delivery → Logging

Trigger: Something starts the process. A form submission, a scheduled time, a new email, a file upload, a webhook from another system.

Intake: The system collects and normalizes the input. Extracts relevant data, handles formatting differences, validates that minimum requirements are met.

AI Processing: The AI analyzes, classifies, generates, or transforms the data according to the logic defined in the spec.

Validation: Automated checks verify the AI's output. Format correct? Data complete? Confidence level above threshold? If validation fails, the item is flagged for human review.

Output: The processed result — a report, a CRM update, a classified document, a notification, a draft response.

Delivery: The output goes where it needs to go. Email, CRM, database, notification channel, shared drive.

Logging: Every step is recorded. What came in, what the AI did, what went out, any errors or flags. Full audit trail.

Three Use Case Scenarios in Detail

1. Lead Qualification Pipeline

Your current process: A potential client submits a form on your website. Someone on your sales team reads it (eventually — maybe the same day, maybe two days later). They look up the company, assess whether the lead is worth pursuing based on experience and gut feeling, enter the information into your CRM, and either schedule a follow-up or mark it as low priority.

With process automation: The form submission triggers the pipeline. The AI scores the lead based on criteria you defined (company size, industry, budget indication, specific needs mentioned). It updates your CRM with the score and reasoning. High-priority leads get an immediate notification to your sales team plus a personalized follow-up email scheduled within 2 hours. Medium leads get a follow-up within 24 hours. Low-priority leads get a polite acknowledgment.

Result: No lead sits in the inbox for two days. Every lead gets a response. Your sales team focuses on the high-value conversations instead of triaging an inbox.

2. Document Processing

Your current process: You receive PDFs from clients or partners — invoices, contracts, reports, applications. Someone opens each one, reads it, extracts the relevant data points, enters them into your system, classifies the document, and files it. For a company processing 50-100 documents per week, this is a part-time job.

With process automation: PDFs are uploaded (or arrive via email). The AI extracts structured data — dates, amounts, names, categories, key clauses. It classifies the document type. Data is entered into your system automatically. A summary report is generated. Documents that the AI cannot process with high confidence are flagged for human review.

The automation does not replace the human for edge cases. A scanned document with poor image quality, a contract in an unexpected format, an invoice missing required fields — these get flagged, not guessed at.

3. Weekly Reporting

Your current process: Every week, someone pulls data from 3-4 different sources (CRM, accounting software, project management tool, spreadsheets). They copy numbers into a report template, calculate trends, write commentary, format it, and send it to stakeholders. Takes 4-6 hours every week. The person who does it hates it.

With process automation: Data is pulled automatically from all sources on a schedule (Friday morning, for example). The AI analyzes trends — what changed, what is notable, what needs attention. It generates a report in your template, with your formatting, including written commentary that matches your style. The report is delivered to stakeholders automatically. A draft is also sent to you for review before final delivery if you prefer.

Six hours per week recovered. Fifty weeks per year. That is 300 hours — roughly two months of full-time work — returned to your team annually.

Error Handling

What happens when the AI is uncertain? This is defined in the spec for every project, but the general pattern is:

›High confidence (above threshold): AI acts autonomously. Output delivered without human intervention.
›Medium confidence (between thresholds): AI processes the item but flags it for human review before delivery.
›Low confidence (below threshold): AI stops, logs the issue, and routes the item to a human. No guessing.

The thresholds are set during specification based on your risk tolerance. A customer-facing system might have higher thresholds than an internal reporting tool.

Every error, flag, and exception is logged. You can see exactly what the system processed, what it flagged, and what it could not handle. No black boxes.

Support Period

30 days of post-launch support for bugs. Same terms as Tier 2: bugs (system does not do what the spec says) are fixed at no additional cost. Feature requests and scope changes are separate work.

What Makes Tier 3 Different From Tier 2

The key difference is autonomy. A Tier 2 assistant helps a human do their job. A Tier 3 automation replaces a manual process entirely for the standard cases. The human only steps in for exceptions and edge cases.

This means Tier 3 projects require more specification work (because the system needs to handle exceptions gracefully), more testing (because the system runs without constant supervision), and more error handling (because problems need to be caught automatically, not by a person watching the output).

Timeline

4-8 weeks from signed specification. Integration complexity is the main variable. If your tools have clean APIs, we are closer to 4 weeks. If we are dealing with legacy systems or unusual data formats, closer to 8. Most Tier 3 projects land at 5-6 weeks.

Tier 4 — AI System (from 10,000 PLN)

What it is: A multi-component AI system designed for a specific business function. This is not a single assistant or a single pipeline — it is multiple AI models working together, each handling a specific domain, coordinated by a central agent.

Think of it as a team, not a single employee. A Tier 2 project is one person doing one job. A Tier 4 project is a team of specialists with a manager coordinating their work.

Architecture

A Tier 4 system typically includes:

›Multiple AI models, each optimized for a different task (classification, generation, analysis, summarization)
›A coordination layer that routes work to the right model, manages dependencies between components, and handles failures
›Shared data infrastructure — knowledge base, vector database, caching, logging
›Integration layer connecting to your existing tools and data sources
›Monitoring and alerting — the system watches itself and notifies you when something needs attention

Three Use Case Scenarios in Detail

1. Knowledge Management System

Your company has 500+ documents across shared drives, wikis, email archives, and the heads of senior employees. Finding the right information takes too long. New employees cannot get up to speed. Institutional knowledge walks out the door when someone leaves.

A knowledge management system handles this end-to-end:

›Intake: New documents, emails, and meeting notes are ingested and indexed automatically. The system extracts key information and categorizes it.
›Retrieval: Anyone in the organization can ask questions in natural language. The system searches across all sources, synthesizes the answer, and cites its sources.
›Q&A: The system does not just find documents — it answers questions. "What was the outcome of the Johnson project in Q3?" gets a direct answer with a link to the source, not a list of 20 documents to read.
›Update workflows: When policies change, the system flags affected documents and answers that need updating. Information stays current.
›Access control: Different teams see different information. The system respects your existing permission structure.

2. Client-Facing AI

Your clients interact with your company through multiple channels — email, web form, phone (transcribed). Each interaction needs to be triaged, responded to, and tracked. For a company handling 200+ client interactions per week, this overwhelms the support team.

A client-facing AI system handles the flow:

›Intake triage: Every incoming interaction is classified by type (inquiry, complaint, request, feedback), urgency, and topic. Routing happens automatically based on rules you define.
›Response generation: For routine interactions (FAQs, status updates, standard requests), the AI drafts responses matching your communication style. A human reviews before sending for high-stakes communications.
›Escalation logic: When the AI cannot handle something (complex complaint, legal issue, angry client), it escalates immediately with full context attached. The human receiving it does not start from zero.
›Feedback loop: Client satisfaction data feeds back into the system. Response quality improves over time based on which responses led to good outcomes and which did not.
›Reporting: Weekly summaries of interaction volume, types, resolution times, satisfaction scores, and trends.

3. Operations Intelligence

Your business generates data from multiple sources — sales, operations, logistics, finance — but nobody has time to analyze all of it. Decisions are made on gut feeling or outdated reports. Problems are discovered late.

An operations intelligence system provides ongoing analysis:

›Data ingestion: Automatic connection to your data sources (databases, APIs, spreadsheets, external feeds). Data is normalized and stored for analysis.
›Analysis: AI models identify patterns, trends, and anomalies. Not just "sales went up" but "sales of product X in region Y increased 23% after we changed pricing, while product Z declined in the same period — here is the likely explanation."
›Anomaly detection: The system alerts you when something deviates from expected patterns. A sudden spike in support tickets, an unusual drop in conversion rate, a supplier consistently delivering late — flagged before you notice.
›Alerting: Configurable alerts based on thresholds and conditions you define. Not a flood of notifications — targeted alerts for things that need attention.
›Reporting: Automated reports at whatever cadence you need (daily, weekly, monthly) with AI-generated commentary explaining what the numbers mean.
›Recommendations: Based on the analysis, the system suggests actions. Not just "sales are down" but "sales are down because of X, here are three options and their likely outcomes."

Team Onboarding Session

Tier 4 projects include a team onboarding session of up to 4 hours. This covers:

›System operation: daily use, monitoring, basic maintenance
›Architecture overview: what each component does and how they work together
›Advanced configuration: changing thresholds, adjusting rules, updating knowledge bases
›Common scenarios: what to do when the system flags something, how to handle escalations
›Troubleshooting: recognizing common issues and resolving them

The session is recorded. New team members can watch it later without needing another live session.

Support Period

60 days of post-launch support for bugs. This includes:

›Bug fixes for issues where the system does not behave as specified
›Configuration adjustments to address unexpected edge cases
›Performance monitoring and optimization during the stabilization period
›Priority response (within 8 business hours for critical issues, 24 hours for non-critical)

What is not included in the 60-day period: new features, new integrations, scope expansion, or changes to the specification. Those are new projects or part of a maintenance retainer.

What Makes Tier 4 Different From "Just Building Multiple Tier 3 Projects"

You could build three separate Tier 3 automations. They would each work independently. But they would not share context, coordinate decisions, or pass information between each other. Tier 4 is for when the components need to work together as a system.

A lead qualification pipeline (Tier 3) qualifies leads. A client communication system (Tier 3) handles responses. But a client management system (Tier 4) qualifies the lead, generates a personalized response based on the qualification, schedules the follow-up based on urgency, and reports on the entire funnel — all as one coordinated process.

The coordination is where the complexity (and the value) lives.

Timeline

8-16 weeks from signed specification. Depends on number of components, integration complexity, and data preparation requirements. Most Tier 4 projects land at 10-14 weeks.

Post-Project: Maintenance Retainer (2,000-4,000 PLN/month)

For clients who want ongoing support after the build is done.

Your system is live, your team is trained, the handoff is complete. But you want someone who knows the system inside and out available for when things change — and things always change.

What is Included

›Prompt updates. When your processes change, your AI prompts need to change too. We adjust them to match your evolving business.
›Model upgrades. When Anthropic releases Claude 5 or OpenAI releases GPT-6, your system can benefit from the improvement. We handle the upgrade, test compatibility, and ensure nothing breaks.
›Integration changes. Your CRM updated their API. Your email provider changed their webhook format. Your database moved to a new server. We handle the technical adjustments.
›Performance monitoring. We keep an eye on system health, accuracy trends, and resource usage. If performance degrades, we catch it before it becomes a problem.
›Priority support. Questions and issues are addressed within 24 hours on business days.

What is NOT Included

New features or scope expansion. Those are new projects with new specifications and new quotes. The retainer covers maintenance and support for what exists, not building new things.

If you need both maintenance and new development, we handle them as separate tracks with separate scopes.

Terms

Month-to-month. Cancel anytime with 30 days notice. No long-term contracts, no cancellation fees.

Typical Clients

Tier 3 and Tier 4 projects where the business depends on the system daily. If your lead qualification pipeline processes 50 leads per day and your revenue depends on it working, you want someone on call who knows the system. That is what the retainer is for.

Tier 2 clients sometimes use the retainer for the first 2-3 months after handoff while their team builds confidence, then cancel when they are comfortable operating independently. That is perfectly fine.

Choosing the Right Tier

Not sure where you fit? Here is a simple guide:

"I want to understand if AI is relevant to my business." Start with Tier 1. It costs 149 PLN and gives you a clear, written answer.

"I have one specific task I want AI to handle." Tier 2. One assistant, one workflow, one integration.

"I have a process with clear inputs and outputs that runs repeatedly." Tier 3. Full automation, end-to-end, with error handling and logging.

"I need multiple AI components working together across a business function." Tier 4. Multi-model system with coordination, monitoring, and team onboarding.

"I am not sure." Start with Tier 1. It is designed exactly for this situation. You will walk away knowing which tier fits your needs, or knowing that AI is not the right tool right now.

"I have a limited budget but a real problem." Talk to us. The Tier 1 consultation (149 PLN) might reveal that a Tier 2 assistant solves 80% of your problem. You do not always need the biggest solution — sometimes the right answer is the simplest one that works.

Upgrading Between Tiers

Projects often evolve. Here is how upgrades typically work:

Tier 1 → Tier 2: The most common path. The AI Guide identifies the best first project, and we move into a Tier 2 build. The 149 PLN from the Guide is deducted from the project cost.

Tier 2 → Tier 3: When the assistant proves its value and you want full automation. The assistant handled email classification well — now you want the entire workflow automated end-to-end, including routing, response drafting, and CRM updates.

Tier 3 → Tier 4: When one automation is not enough and you need multiple components working together. The lead qualification pipeline works great — now you want to connect it with client communication, reporting, and operations monitoring.

We design for extension when expansion is likely. If you tell us during Tier 2 that you are probably going to want Tier 3 later, we architect the assistant so it can become part of a larger pipeline without rebuilding from scratch. We note this in the spec.

Important: upgrading does not mean the previous tier was wasted. The Tier 2 assistant becomes a component of the Tier 3 automation. The Tier 3 pipeline becomes one arm of the Tier 4 system. Each tier builds on what came before.

What We Do Not Build

Some requests fall outside what we do. In each case, there is a reason.

Systems designed to deceive users about their AI nature. Trust is the foundation of any customer relationship. If your customers discover the AI was pretending to be human, you do not just lose their trust — you lose them. Permanently. We will not build something that sets that trap.

Autonomous trading or financial decision systems. Financial markets require human judgment for consequential decisions. AI can analyze data, identify patterns, and make recommendations. But the human decides and the human executes. The moment you remove human oversight from financial decisions, you are one edge case away from catastrophic loss.

Surveillance or monitoring systems without employee knowledge. It is illegal in the EU under GDPR and labor law. And it is ethically wrong everywhere else. If you want to monitor employee productivity, do it transparently with their knowledge.

Projects without written specifications. Unspecified projects always fail. Not most of the time — always. 100%. No exceptions. If the scope is not written down, both sides have a different picture in their heads, and the result satisfies nobody. We will not start building without a spec.

Retainer arrangements with undefined scope. "Just fix whatever comes up" leads to scope creep, burnout, and bad outcomes for both sides. If the scope is not defined, neither is "done," and a project that is never done is a project that is never successful.

We also do not do:

›"Just a small thing" additions to completed projects without a new spec and agreement
›Hourly billing without a scope boundary
›NDAs that prohibit us from describing the category of work we performed

On Pricing

Prices listed are starting points. The final quote depends on integration complexity, number of data sources, revision cycles, and specific requirements identified during Discovery.

We quote after Discovery, not before. Sending a price without understanding the problem is guessing, and we do not guess with your money. Discovery is free precisely so we can give you an accurate quote instead of an optimistic one.

We do not negotiate scope to fit a budget. We scope to fit the problem, then price accordingly. If you tell us your budget is 3,000 PLN and the problem requires 6,000 PLN of work, we will tell you that directly. We might suggest a smaller scope that fits your budget and solves the most valuable part of the problem, but we will not pretend the full solution costs less than it does.

Payment structure:

›Tier 1: Full payment upfront (149 PLN).
›Tier 2 and above: 50% on specification approval, 50% on delivery.

The 50/50 structure protects both sides. You do not pay the full amount before seeing the build. We do not do the full build before receiving a commitment.

If Discovery reveals the project is outside your budget, we say so before the spec phase begins. No wasted time. No drawn-out discussions where we both know the numbers do not work. Honest, direct, and early enough that nobody has invested significant effort.

Ongoing model costs: Our project fees cover the build. Running the system requires API calls to model providers (Anthropic, OpenAI), which are billed directly to you by those providers on a usage basis. Typical monthly model costs range from $30-$500 depending on volume. We estimate these during Discovery so there are no surprises.

Currency: All prices are listed in PLN (Polish zloty). For international clients, we invoice in EUR or USD at the current exchange rate at the time of invoicing.

VAT: Prices listed are net (before VAT). Polish clients add 23% VAT. EU clients with a valid VAT number pay reverse charge. Non-EU clients pay no VAT.

The Consultation FAQ